Using output probability distribution for improving speech recognition in adverse environment

This paper proposed a method to improve the accuracy of small vocabulary isolated word speaker-independent speech recognition in adverse environment. The proposed approach is implemented by using Output Probability Distributions (OPDs) and Support Vector Machine (SVM). OPDs improve the system performance by modeling inter-word relationships; then SVM classifiers are used to discriminate the difference between OPD models. The system was tested using isolated Mandarin digits database, corrupted with the NOISEX-92 database. The experiments have achieved good result in noise conditions, the WER dropped about 30% on average when compared to the HMM recognizer.